New AI model for breast cancer treatment plans

ENGINEERINGNET.BE – When developing breast cancer treatment plans, there is still a lot of manual work in creating accurate radiographic plans and mapping out individual patient organs.

says Quinn Horkmans, clinical physicist at Katharina.

The simulator contains a model of treatment equipment and images of an individual patient’s anatomy. An experienced laboratory technician then develops a radiation plan.

This work can be automated. Artificial intelligence is a practical and effective way to record, calculate and interpret knowledge and experience. Put dozens of pre-made plans into a template and train the computer to make them yourself.

PDeng student Nienke Bakx from Eindhoven University of Technology evaluated two software models for this purpose: the open source U-Net model, which is based on Convolutional Neural Networks, and CNNs. and the CARF model, where cARF stands for Atlas Regression Forest.

Bakx filled out both forms with data from more than a hundred patient treatment plans. The trained models were tested during a clinical trial with data from twenty new patients.

Radiation therapists and lab technicians created their plans for these patients manually and compared them to automatically generated plans to further improve the program. The better the input, the better the output.

95% of what came with the computer turned out to be usable without any manual modification. The U-Net model scored slightly better than the cARF model, prompting Swedish partner RaySearch to use this model in RayStation’s patient planning program for radiotherapy.

A fully trained and validated business model will be published in clinical practice from May. By the way, radiation therapists and lab technicians check everything set up automatically.

The model is also ready for automatic mapping of organs and glandular regions, and a clinical trial is now underway.

In the above figure: two axial cross-sections of a female patient with breast cancer, showing the limited breast, lymph nodes, lungs and heart.